Artificial neural network in predicting rice yield

Rice production is one of the major sectors that play an important role on the national economy. Hence, site specific nutrient management is crucial for a sustainable agriculture. Therefore, precision agriculture and information technology is really important to balance crop productivity. The application of neural network to the task of predicting crop yield is essential. The objectives of this paper were to: 1) investigate whether artificial neural network (ANN) model could predict rice yield based on soil parameters; 2) determine the most affected soil properties towards rice yield; 3) compare the effectiveness of multiple linear regression model to ANN. Models were developed using historical data collected in Block C, Sawah Sempadan, Selangor, Malaysia for two continuous seasons. Season 1 is dry season while Season 2 is wet season. External factors such as weather, farmer’s practices etc. were not being considered in this study. ANN showed more accurate results than regression model. ANN model resulted in r2 of 0.71 and 0.69 for Season1 and Season 2 respectively. While in linear regression, r2=0.12 and 0.02 for Season1 and Season 2 respectively. The results show that ANN model is more reliable than regression model in predicting rice yield. It can be conclude that ANN model is simple yet accurate.

PDF _TechnicalPapers_CAFEi2012_74.pdf Restricted to Repository staff only Download (50kB) Official URL or Download Paper: http://cafei.upm.edu.my/download.php?filename=/Tec...

Actions (login required)

View Item